1. Your V register in MATHR is actually an IEEE floating point DWORD result, so it utilizes TWO V locations. Make sure you do not have ANY OTHER INSTRUCTIONS WRITING to the V location or to the NEXT V location (e.g. if MATHR V2000, make sure no instruction is using V2000 and V2001). Remember V's IDs are octal, so if MATHR V2007, then it is V2007 and V2010, not V2008.
2. Make sure the input logic driving the MATHR instruction enables the MATHR, e.g. stick an INC box below the MATHR box and it should increment the V register in the INC box - if it is NOT incrementing, then your MATH box is not enabled either.